Aug 2026· Journal of Intelligent Decision Making and Information Science· Vol 3, pp. 1174-1197· 0 citations· 29 references
TL;DR
The proposed CTAE framework provides a robust and trustworthy explainability solution for high-stakes multimodal AI applications requiring transparent and cognitively reliable decision interpretation.
Abstract
With the growing use of multimodal AI systems in healthcare applications, conversational AI, affective computing, and decision support systems, there is an increased demand for explainable and trusted AI algorithms. But current methods used to explain AI systems typically result in attribution-based explanations without any causal reasoning abilities, perturbation resistance, and consistent explanations across different modalities. This paper presents a new method called Counterfactual Trust-Aware Explainability (CTAE) that aims at developing an explainability framework for multimodal AI decision systems. Our approach integrates multimodal counterfactual reasoning, cross-modal consistency, perturbation resistance, trust calibration, and human-oriented evaluation into a unified explainability framework. Evaluation of the framework was conducted on the MELD, IEMOCAP, and CMU-MOSEI benchmark datasets based on measures like explanation fidelity, perturbation stability, contradiction rate, semantic consistency, causal alignment, and human trustworthiness perception. The experimental results showed that the CTAE framework yielded greater scores in explanation fidelity (0.892), perturbation stability (0.874), and trustworthiness (4.61) when compared to methods like SHAP, LIME, attention visualization, and traditional counterfactuals. Additionally, CTAE demonstrated lower contradictions and better performance in flipping decisions by constraint-based perturbations. Finally, human-centered evaluation of the explanations generated by the CTAE framework confirmed improved quality, usability, and trustworthiness of the explanations across multimodal interaction settings. Overall, the proposed CTAE framework provides a robust and trustworthy explainability solution for high-stakes multimodal AI applications requiring transparent and cognitively reliable decision interpretation.
Explainable artificial intelligence (XAI) has become a central methodology for developing transparent, accountable, and human-centered AI systems. As data-driven models are increasingly deployed in high-stakes and socially consequential settings, explanations are expected not only to illuminate model behavior but also to support validation, error analysis, fairness auditing, regulatory compliance, and effective human–AI collaboration. This Editorial introduces the Special Issue “Explainable Artificial Intelligence Technology and Its Applications” and situates its contributions within the broader trajectory of XAI research. We briefly review major methodological families, including intrinsic interpretability, local surrogate and Shapley-value explanations, gradient- and perturbation-based visual attribution, counterfactual and causal explanations, and human-centered evaluation. We then highlight representative contributions in this Special Issue, which demonstrate how XAI is moving from generic explanation visualizations toward domain-sensitive, data-aware, and operationally reliable methods in network security, computer vision, knowledge graphs, and financial decision support. Finally, we discuss future directions, emphasizing faithful and plausible explanations, causal and multimodal reasoning, real-time and hardware-efficient deployment, trustworthy governance, and the emerging role of XAI in education.
Xuewen Sun, Lan Tian, Weidong Zhou et al.· Applied Sciences· 0 citations
Concept-based explainable artificial intelligence (AI) can make model reasoning more human-understandable, but concept-level outputs are not automatically trustworthy. We introduce ConceptSMILE, a model-agnostic perturbation-based auditing framework for evaluating the reliability of concept-based explanations. Rather than replacing SMILE, ConceptSMILE extends its perturbation-based logic from feature- or region-level attribution to the auditing of human-understandable concept explanations. The framework perturbs input regions, measures concept-response shifts, applies locality weighting, and fits an XGBoost surrogate to approximate local concept behaviour. Reliability is assessed through attribution accuracy, surrogate fidelity, faithfulness, stability, and consistency. We evaluate ConceptSMILE on retinal fundus images by comparing MedSAM-derived visual concepts with VLM-based semantic concepts. Results show that reliability varies across concepts and pathways: MedSAM achieves stronger spatial attribution and the highest surrogate fidelity ($R^2 = 0.8503$, $R_w^2 = 0.8465$), while the VLM pathway shows stronger vessel faithfulness and stronger stability under selected artefact conditions. ConceptSMILE provides an independent audit layer for evaluating the trustworthiness of concept-based XAI.
Mohadeseh Mollapour, K. Aslansefat, Zeinab Dehghani et al.· 0 citations
Artificial Intelligence (AI) has transformed modern communication systems by enabling intelligent interactions, automated content generation, personalized recommendations, real-time translation, and conversational support. The widespread adoption of AI in areas such as social media, healthcare, education, customer service, and enterprise communication depends largely on user trust. Trust in AI communication systems is influenced by factors such as transparency, reliability, explainability, privacy, fairness, and accuracy. However, concerns regarding bias, misinformation, data privacy, and lack of explainability can reduce user confidence. This study examines technological, psychological, and social factors affecting trust in AI-based communication systems. The findings reveal that transparency, reliability, privacy protection, and explainability are critical for building long-term user trust. The study concludes that trust is a multidimensional concept involving technical performance, ethical considerations, and user experience, highlighting the need for trustworthy AI systems to support effective and responsible digital communication.
Michael Rabin· International Journal of Inn...· 0 citations
It is argued that calibration-aligned design (rather than trust maximization alone) should guide the development and assessment of high-stakes AI decision support, because reductions in reported trust do not consistently translate into commensurate changes in reliance behavior.
Two short vignettes and a design guide to help human factors researchers create explanation systems that are practical and trustworthy are presented to show how explainability can become part of the care system instead of being treated as a separate technical feature.
T. Mamun, Laurie Novak, M. Salwei· Proceedings of the Internati...· 0 citations
Artificial intelligence is increasingly used to support clinical decision making, yet concerns remain regarding algorithmic aversion, automation bias and the preservation of meaningful human oversight; while explainable AI aims to improve transparency, less attention has been devoted to the design of human–AI interaction protocols. This study investigates Frictional AI, an interaction paradigm that introduces cognitive friction to encourage critical engagement with AI recommendations. First, semi-structured interviews were conducted with a legal expert and a psychologist and analyzed through thematic analysis to identify legal, ethical, and cognitive requirements for AI-assisted decision support. Second, a user study involving 96 medical residents compared three interaction protocols: a conventional explainable AI-first design (XAI) and two friction-based protocols, namely a judicial protocol based on juxtaposed explanations (Judicial AI, JAI) and an adjunct protocol requiring an initial unsupported decision before AI exposure (AAI). Diagnostic accuracy and confidence, perceived usefulness, completion time, and reliance patterns were evaluated. The interviews highlighted the importance of human-centered explanations, contrastive reasoning, preservation of professional responsibility, and the role of user studies in evaluating human–AI interaction. The quantitative results showed that none of the AI-assisted conditions improved diagnostic accuracy relative to the no-support baseline. However, JAI achieved performance comparable to the baseline, outperforming XAI and AAI, and exhibited the lowest level of over-reliance. Overall findings suggest that the effectiveness of decision-support systems depends not only on model performance and explanation quality but also on interaction design. In conclusion, while preserving diagnostic performance, judicial protocols showed promise in mitigating automation bias and promoting active cognitive engagement in clinical decision support.
Samuele Pe, Laura Bergomi, G. Nicora et al.· Machine Learning and Knowled...· 0 citations